Adaptive algorithm based on least mean p-power error criterion for Fourieranalysis in additive noise

This work presents a novel adaptive algorithm for the estimation of discret e Fourier coefficients (DFC) of sinusoidal and/or quasi-periodic signals in additive noise, The algorithm is derived using a least mean p-power error criterion. It reduces to the conventional LMS algorithm when p, takes on 2, It is revealed by both analytical results and extensive simulations that t he new algorithm for p = 3, 4 generates much improved DFC estimates in mode rate and high SNR environments compared with the LIMS algorithm, whereas bo th have similar degrees of complexity. Assuming the Gaussian property of th e estimation error, the proposed algorithm, including the LMS algorithm, is analyzed in detail. Elegant dynamic equations and closed-form noise misadj ustment expressions are derived and clarified.